Add dataset card for COP-GEN

README.md

@@ -0,0 +1,41 @@
+---
+task_categories:
+- any-to-any
+---
+
+# COP-GEN: Latent Diffusion Transformer for Copernicus Earth Observation Data
+
+COP-GEN is a multimodal latent diffusion transformer that models the joint distribution of heterogeneous Earth Observation (EO) modalities (optical, radar, elevation, and land-cover) at their native spatial resolutions. By parameterising cross-modal mappings as conditional distributions, COP-GEN enables flexible any-to-any conditional generation, including zero-shot modality translation without task-specific retraining.
+
+- **Project Page:** [https://miquel-espinosa.github.io/cop-gen/](https://miquel-espinosa.github.io/cop-gen/)
+- **GitHub Repository:** [https://github.com/miquel-espinosa/COP-GEN](https://github.com/miquel-espinosa/COP-GEN)
+- **Paper:** [COP-GEN: Latent Diffusion Transformer for Copernicus Earth Observation Data](https://huggingface.co/papers/2603.03239)
+
+## Dataset Summary
+
+This dataset is used as part of the COP-GEN project. It includes a stochastic benchmark built from multi-temporal Sentinel-2 observations, Sentinel-1 radar data, and digital elevation models (DEM), enabling distribution-level comparison of generative EO models beyond single-reference, pointwise metrics.
+
+## Sample Usage
+
+To reproduce the distribution-level results reported in the paper using the stochastic benchmark, you can run the following commands from the official repository:
+
+```bash
+pip install -r benchmark/stochastic/requirements.txt
+python -m benchmark.stochastic.run --output metrics.csv
+```
+
+Note: The first run downloads approximately 72 GB from Hugging Face and caches it locally.
+
+## Citation
+
+```bibtex
+@article{copgen2026,
+    title   = {COP-GEN: Latent Diffusion Transformer for Copernicus Earth
+               Observation Data},
+    author  = {Espinosa, Miguel and Gmelich Meijling, Eva and Marsocci,
+               Valerio and Crowley, Elliot J. and Czerkawski, Mikolaj},
+    year    = {2026},
+    journal = {arXiv preprint arXiv:2603.03239},
+    url     = {https://arxiv.org/abs/2603.03239},
+}
+```